Raven's Progressively More Annoying Matrices

[sixes_and_sevens]

I'm currently torturing myself with Malcolm Gladwell's most recent pop-psych outpouring Outliers, in which he mentions Raven's Progressive Matrices, and provides an example of a particularly tricky one:



Annoyingly, in the book the answer is immediately provided, but that wasn't a great deal of immediate help in figuring it out.

Spoiler:

It's A, allegedly.

I think I've found a pattern that's consistent with that answer, but it's a bit of a stretch.

Spoiler:

Each member of the series has one non-central intersection of a row and a column in which no symbols repeat. The only two answers with non-repeating intersections are A and B. The pattern starts to stretch with the "non-central" constraint, which is pretty arbitrary. There are nine series members, so a more aesthetic resolution to the pattern would be for each intersection to appear in one of the nine squares in each grid, and the correct answer to be the one remaining unoccupied grid, but reading from left to right, members 4 and 8 have their intersection in the same position. It's about this point at which I begin to feel like I'm clutching at straws.

Anyone else fancy hatching some pet theories?

(Oh, I'm also a bit new, so whilst I think this is appropriate for this part of the forum, if it's somehow not, please blatt it with my blessings).

[Qaanol]

For communication purposes, I'll refer to locations in the telephone-keypad pattern,
123 (row 1)
456 (row 2)
789 (row 3)

Some initial observations: In row 2 of the puzzle, row 1 of each matrix does not change. Matrix 3 rotates 90° counterclockwise to become matrix 4. Matrix 6 rotates 90° the same way to become matrix 7. That's all I see so far.

[Excalibur0998]

Have you considered...

Spoiler:

That it's just simple transformations?

I found an example:

Spoiler:

The top right square and the center left square are 90 degree rotations. Same with center right and bottom left.

Edit: I got beat'd.

[howardh]

This is probably a bit of a stretch too but it's consistent with two columns of the six matrices to the right

I'll use the same numbering system as Qaanol to label each matrix

Spoiler:

Look at the third column of matrix 2, 5, and 8
the pattern is:
♣
♦
♣
♦
♣
♦
♥
♦
♣
Compare that to the second column of matrix 3, 6, and 9
♥
♣
♥
♣
♥
♣
♦
♣
♥

Similar patterns are in the second column of matrix 2, 5, 8 and the first column of matrix 3, 6, and 9

[sixes_and_sevens]

I am so horribly, horribly close.

Spoiler:

Give hearts value 1, clubs value 2 and diamonds value 3. Drop each symbol by one value (diamonds to clubs, etc., hearts wrap back round to diamonds). Then shift the entire matrix left by one position (position 1 becomes position 9, 2 becomes 1, etc.). This transformation explains transformations 2->3, 4->5, 5->6, 7->8 and 8->9. Transformations 3->4 and 6->9 are a 90 degree anticlockwise rotation.

I can't account for 1->2 in any way, shape or form. There is no orientation of matrix #1 which results in the pattern XXYYXZZZY, so no amount of cycling through symbols will account for it.

[Jonoleth]

Spoiler:

There also seems to be a pattern in the diagonal symbols of the corresponding diagonal positions. Symbols 3, 5 and 7 in positions 3, 5 and 7 all have a X, X, Y pattern, where all three symbols are represented. Same thing with 1, 5 and 9 in 1, 5, A.

It's like this thing just has a ton of unrelated little patterns in it just to fuck you over

[Another]

Original source of this puzzle was named Outliers and I suspect that it is a hint. Specifically:

Spoiler:

1->2 transformation is an outlier. Ignore.

In this case it really messes with your expectations of usual logical puzzle. Two types of puzzle in one - "find one that doesn't belong to a set of elements" and "find next element in a sequence".

[wshanley]

Spoiler:

The second and third columns of boxes 2,5,8 become the first and second columns of boxes 3,6,9 under the transformation H->D,D->C,C->H. This eliminates all answers but A.

Reminiscent of http://xkcd.com/153/

[suzy]

I don't know where this puzzle originally came from but it is not Raven. I have the Advanced Progressive Matrices Set I and Set II and that puzzle is not in there. All of the Raven items can be solved with a rule or combination of rules so that each part of the set belongs. There are no items where one part does not follow the same rule or rules as the other parts.

[sixes_and_sevens]

I guess that's another entry in the list of complaints I have for Malcolm Gladwell if I ever meet him, then.

[coryalantaylor]

Ok seriously though. I spent over an hour figuring this one out before I read the answer. I was 100% sure it was H. Is A. actually right? or was it just a typo?

[GeorgeH]

A is correct:

Spoiler:

Step 1
Look at each matrix individually, and assign it a number based upon the number of shapes that have no neighbors of the same shape; your result will be:

1-3-2
2-1-3
3-2-?

Clearly, we need a "1" to fill the pattern.

Step 2
How many different ways are there to write a "1"? Only 3:

Method 1)
--*
*--
*--

Method 2)
--*
*--
-*-

Method 3)
*-*
*--
---

Looking the two "1"'s already in play, we see that we have a "Method 1" and a "Method 2", but no "Method 3". Obviously this lack of completeness cannot be tolerated, so we need a "Method 3" "1" to complete the pattern.

Conclusion
We circle answer "A", because it's the only "Method 3" answer available. The specific shapes, orientation of the "1" pattern, and relationships between the "1" pattern and the "2" and "3" patterns are left as an exercise for the obsessive. Personally, if the puzzle maker couldn't be bothered to hide the answer better (and/or eliminate alternate solutions) I can't be bothered to care.

[speedyld8]

Hello friends, I must confess that I'm horrible at maths yet was able to solve this one using a bit of spatial logic.

Spoiler:

It has to do with the contiguousness of the suits. Each suit has a ratio of contiguous to non-contiguous instances which holds across the row.

Ratio 1: 2 x non-contiguous diamonds, 1 x contiguous diamonds
Ratio 2: 2 x contiguous hearts, 1 x non-contiguous hearts
Ratio 3: 2 x contiguous clubs, 1 x non-contiguous clubs

To find the missing 3x3 box we have to look at each suit individually across the row to see what is missing from each ratio and find a box which enables each of the ratios described above.

Based on the two given boxes, we can see that the box that will complete the set should have the following:

• Contiguous diamonds
• Non-contiguous hearts
• Contiguous clubs

Choice 'A' is the only one of the given possibilities which satisfies all of these requirements. Indeed, choice 'A' is the only possibility with non-contiguous hearts.

[Stolten]

Hello

I would like to add a similar puzzle to this thread or to the logic puzzle forum. How do you make attachments? Anyone can help?

Best regards, Stolten

[Stolten]

Here is a similar puzzle that perhaps could be solved with the same rules as speedyld8 or GeorgH provided. Anyone care to give it a try?

Best regards, Stolten

[Stolten]

No answer yet so I will give the solution that I came up with:

Spoiler:

Look at each matrix individually, and assign it a number based upon the number of shapes that have no neighbors of the same shape; your result will be:

1-3-0
0-3-4
4-1-?

I do not recognize a pattern in the matrix but if you add the row elements of the matrix you will get the column matrice

1+3+0 = 4
0+3+4 = 7
4+1+? = 5 + ?

The elements of the column matrice reseble a numeric serie where each element can be denominateed x(n)
The serie
x(n+1)=x(n) + n + 1
With the initial value
x(0)=2

Would yield

x(1)= 4
x(2) =7
x(3) = 11

As x(3) = 11 the question mark (?) in our third row element of the column matrix should be 6 because 5 + 6 = 11

We thus find that the only solution that comply with this demand is answer A.

To strenghten our proof we can sum the column elements in the first matrix that now looks like this

1-3-0
0-3-4
4-1-6

We get the row matix

5-7-10

The elements of the row matrice resemble another numeric serie where each element can be denominateed x(n)
The serie
x(n+1)=x(n) + n
With the initial value
x(0)=4

Would yield

x(1)= 5
x(2) =7
x(3) = 10

Thus both the sum of the rows and the sum of the columns follow numeric series and we should be confident that A is the correct answer to tha puzzle.

Anyone has another solution for the puzzle?

Best regards, Stolten

[speedyld8]

I have the answer, with some explanation for it:

Spoiler:

It's C

The answer lies in the fact that the columns in each 3 x 3 successive matrix shift one column to the right, and in doing so, the shapes change each time according to the following rule:

squares become circles
circles become triangles
triangles become squares

To check this, apply the transformation rule to column 1 of any matrix in a row and you will see that it is identical to column 2 of the following matrix; the same can be done to column 2 and checking it against column 3 of the following matrix. They will be identical.

The problem of the “lost” column at furthest left of each matrix is resolved thusly:

According to the patterns in the matrices, there are always three of each shape. Thus, the missing shapes necessary to complete each individual shape set are used to create a new first column in each successive set.

If we apply the transformation to the last or rightmost of each matrix column and “recycle” it to the first column of the next matrix, we have the shapes to complete the sets.

However, the final question remains about how to determine the order of the shapes in this first column.

This too follows a hard and fast rule. As each you apply the transformation to the shapes of the rightmost column and “move” it to become the first column of the next matrix, you also shift each shape one row down, with the last row of the individual column getting “recycled” to the top. You can also think of it as switching the first and the third positions in the column. This switching of rows only applies to the single column that is “moved” off of the end of one matrix and onto the beginning of the next.

You will find that these rules apply in all instances, ultimately producing answer C.

N.B. As in the last puzzle posted on this thread, there is also a continuity between each row of matrices – the first matrix in each row is identical to the last matrix in the previous row rotated 90 degrees clockwise. This is very similar but not the same as the continuity between matrix rows in the last puzzle (in which the aforementioned matrices were rotated anticlockwise).

Cheers for this puzzle, Stolten. If you've got more, keep them coming!

[GeorgeH]

No answer yet so I will give the solution that I came up with:

I'm fairly confident that with some effort you could use the same procedure (map a matrix pattern to a number, then take a subset of those numbers and find a mathematical formula that uniquely combines them with one of the answer matrices) to conclude that every answer is uniquely correct. It's still a fun solution, though.

I have the answer, with some explanation for it:

Bastard, you beat me too it.

As a consolation prize, I’ll point out an alternate solution I found (kind of a subset of the correct solution) that someone might arrive at before the correct one:

Spoiler:

It's (actually not) G.

In any column or row, there are only 4 unique patterns:
1) XXX
2) XXY
3) XYX
4) XYZ
(Ignoring YXX as nothing but a reflection of XXY)

Putting the patterns in a (row, column) pair, the given matrices are as follows:
(444, 234) (233, 423) (243, 242)
(242, 342) (423, 334) (213, 433)
(433, 312) (224, 231) (???, ???)

Numbering the collection of matrices as:
1 2 3
4 5 6
7 8 ?

We see that it’s possible to go from 1 to 8 by swapping columns and rows between matrices:
234<>423, 423<>243, 243<>342*;
342<>234, 334<>433, 213<>312*;
312<> 231, ???<>???, ???<>???
*One of two possible transformations

Continuing as best we can, the answer needs to have some permutation of a “123” pattern or a “224” pattern.

The potential answers are as follows:
A – (434, 313)
B – (343, 432)
C – (223, 123)
D – (234, 343)
E – (422, 123)
F – (233, 324)
G – (324, 242)
H – (212, 343)

Therefore the only possible answers are C, E, and G.

If the rules have been explained to you, after casting about for a bit you’ll see that you haven’t got quite the right approach, then notice the “correct” behavior as you examine how the columns transform after eliminating “3” as being redundant (both XXY and XYX have two Xs, XXX as three, XYZ only one.)

If the rules haven’t been explained to you, most importantly that the interval [1,8] is supposed to be linear (…,0,1,2,3,4,5,6,7,8,9…) and not cyclic (…,9,1,2,3,4,5,6,7,8,9,1,…)** you might conclude that G is the only logical choice, as it is the only answer that can transform back to 1. However, if you try to justify the shapes and/or the exact transformation procedure (which row/column goes where) you won’t find a nice and simple rule, which is why C is more correct.

**Dear math nerds: Yes, I know that the integers form an (or the) infinite cyclic group under addition. Let me be colloquial for one damn second.

[Stolten]

Nice solution speedyld8 and thanks for you subset GeorgeH

Here is a less complex matix puzzle. Care to give it a try?

Best regards, Stolten

[GeorgeH]

My answer would be:

Spoiler:

8.

If you take a group of 3 matrices going along a row, it looks like you can connect them together to form an “∞” shape by joining “neat” sides (the outermost edge of the triangle formed by the two lines and the side of the square or the interior defined by the lines when they don't cross; in the 1st matrix it would be the left and right side, the 6th would be the top and right sides, etc.) Note that this is by eyeballing things, I haven’t bothered to load the images into an editing program to see if they match perfectly. Matrix 8 looks to be the only one that could finish the job on the last row, so I select it as my answer.

There could be a more complex pattern going on, but for now I’m happy with 8.

[jestingrabbit]

I'd say

Spoiler:

5

If you linearly order the boxes, the endpoints of the lines in the previous box are the starting point for the lines in the nest box. The only one that has that property is 5.

[Xanthir]

Nope, the answer is

Spoiler:

5. Taking each matrix in order (left to right, then top to bottom), the adjacent pairs of matrices can be hooked together such that the line endpoints on their adjoining sides exactly match. From 1 to 8, the pattern formed looks like this:

Code: Select all

123
  4
  5
  678


And the last matrix needs to thus match its left side against 8's right side. The only one that does so is 5.

I can see the appeal of answer 8 - the spacing between the two lines on its left side is right, but the positioning isn't. What's more, *its* right side looks like a possible match for block 1's left side, giving the whole thing a nice symmetry. But the positioning kills it.

[speedyld8]

Ok, I hadn't checked this thread in a while and didn't see you had put up a new one, Stolten. Looks like I was beaten to the punch on this one! Still, here's my answer, with explanation.

Spoiler:

I arrived at 5 as well, but in a different manner from that of jestingrabbit and Xanthir before me.

To solve this we have to look at the line segments comprising the walls of each square. You will notice that in each and every square there are two walls comprised of single line segments in which any intersecting lines within the square are terminal at the endpoints of the line segment; the remaining walls are comprised of two or more line segments, with the wall being divided by endpoints of the lines within the squares. I'll refer to these types of walls as ‘A’ and ‘B’, respectively.

(Note: it does not matter how many segments are in a type ‘B’ wall. For the pattern it only matters that it is not comprised of a single line segment.)

What we are looking at is the progressive relationships of these kinds of walls to each other within each square across the rows. We see that we only have two basic types of squares, a square in which each type of wall is opposite the other of its kind, or a square in which each type of wall is adjacent to the other of its kind. Among these two types, the orientation of the squares produces six total possibilities.

Visually:

Square Type 1 (opposite):

Type 1a:

---A---
B-----B
---A---

Type 1b:

---B---
A-----A
---B---

Square Type 2 (adjacent):

Type 2a:

---A---
A-----B
---B---

Type 2b:

---A---
B-----A
---B---

Type 2c:

---B---
B-----A
---A---

Type 2d:

---B---
A-----B
---A---

Based on this, we can easily see the top-level pattern across the rows:

Type 1 – Type 1 – Type 2

Thus we can narrow down our answers to choices 3, 4, 5, and 6, as they are the only Type 2 squares in among the possible answers.
Now to find the right answer among these four, we have to further look for a further pattern in the set. For this we look to the subtypes. This produces the following pattern across the matrix:

Type 1a – Type 1a – Type 2b
Type 1b – Type 1b – Type 2d
Type 1a – Type 1a – ???

From this we see that only a square matching Type 2b will do to finish the pattern. Choices 3 and 6 are of Type 2a and choice number 4 is of Type 2c. Choice number 5 is the only one that fits Type 2b out of the four possibilities listed prior.

Some notes:

Type 2d is not represented in any of the choices given so we do not have to rule it out but could do so using the rule exemplified in row 2 of the matrix.

As regards Type 2a and Type 2c, we can rule them out on the basis that we have the rules governing the same subtypes in a row (ie. 2 x Type 1a in a single row and 2x Type 1b in a single row), and we can assume that the rules governing differing subtypes (ie. Type 1a and Type 1b in a single row) will produce these other Type 2 subtypes, and further that the order of the Type 1 subtypes of across the row would differentiate which (2a or 2c) is ultimately produced as the third element.

[Stolten]

Nice solutions. I would have solved it like jestingrabbit and Xanthir.

Here is another one quite tricky one. Anyone care to try it.

Best regards, Stolten

[speedyld8]

This seems simple enough, unless I'm deceived.

Spoiler:

The answer is 6.

Moving rightwards, we see that there is a repeating pattern of columns: a column having two squares followed by a column having one square followed by another column with two squares, and so on.

Moving from a column with two squares to the next, we see that the columns containing two squares combine the patterns of their squares to form the one square in the subsequent column. The operational rule is that the bottom square of the two becomes the top half of the single square in the following column, and the top square becomes the bottom half of that single square.

Now for the progression to continue, there needs to be a column with two squares following the column with one square. There is a rule applicable here as well which allows us to logically get the squares needed in the next column. The rule is that the bottom square of the subsequent two square column is identical with the single square in the preceding column, and the top square is the reflection of the square in the preceding column across its horizontal axis.

To get the missing piece, we just apply the rule going from a two-square column to a one-square column, which produces answer 6. To verify it, we apply the rule going from a one-square column to a two-square column. The only possibly tricky bit that might throw one off is the fact that the final column in the series shows two of the same square, which obfuscates the fact that any sort of operation has taken place; however, it makes sense once we grasp that the single square that precedes this column is a "visual palindrome" across its horizontal axis.

[cba]

I would have said that as well with pretty much the same logic...

[Reinholt1]

new member here. I'd like to go back to the original discussion - on the "Raven's" (whether it is or not) example in Gadwell's book. I dunno about your books. But in my book, the example given here IS NOT the example given in my book. In my book the upper left square of the puzzle and the "H" answer squares are reversed. So in my book (First edition, November 2008) the puzzle is: (D=diamond, C=club, H=heart)

DCC HHC DHH
DDC CHD DCC
HHH DDC CHD

HCD HCD HCD
HCH HDC CHH
DDC CHD DCD

DHD DCH
CHC DHD
HCD HCC


And the H answer given in my book is:
CCD
DHC
DHH


So I guess the question is: which is the typo? My book or the example given here?!?

[jdogg0075]

After doing this problem tonight. I found the quickest/simplest method for me.
Hypothesis:

Spoiler:

There are three independent symbols. Thus a pattern will arise from a three sided geometric object (triangle) with some defined order. This order for me was Heart, Diamond, Club. Remember it must form a three sided triangle.

Process/Answer:

Spoiler:

What you will then notice after seeing/drawing all the possible triangles formed from my hypothesis. You notice the pattern that arises is as such: Top row: 4 triangles each. Middle Row: 5 Bottom row: 6. The answer is the pattern that has six triangles following the same form. A is the only one that has 6.

Let me know what you guys think! I was very excited for this was my first raven pattern. Cheers.

-Justin

[Awoken]

In responce to the first matrices that was given, I came up with G.

The logic behind the choice is about 50% proof,an you're left with two possible answers "G & H" and need a little encouragement to select G over H.
Each of the Matrices has three unique patterns of three showing.

All 8 matrices display 24 unique arrangements of a pattern of three. No pattern of three has a repeating pattern of three in anyother matrices. Answers A - F, A has one repeating pattern in matrices 5, B in 5, C in 1, so on and so forth. F repeats its pattern of three three times. So you're left with G & H. I think most people would select G over H simply because H shows a pattern of three on the same row, However, because H is a possibility and no logical cord can distinguish between G and H the logic isn't full-proof.

[Anubis]

I hate these things, because you can find a pattern that generates any one of the possible answers, so which one is "right" is entirely arbitrary. I wouldn't mind so much if they didn't actually put them on IQ tests and such. Or at least gave you some quantitative metric by which to judge possible solutions.

[kainidge]

I came up with the answer of A

Spoiler:

My Theory was this....
Same working out as howardh

Pattern of matrix 1 is...

center across, diamond heart club, right down, diamond heart club, linking is club

matrix 2 as follows,

center across, club heart diamond, left up, diamond club heart, linking is club

matrix 3 as follows,

bottom across, club heart diamond, right down, club heart diamond, linking is diamond,

matrix 4 as follows,

top across heart club diamond, right down, diamond heart club, linking is diamond

matrix 5 as follows,

top across, heart club diamond, center down, club diamond heart, linking is club

matrix 6 as follows,

left down, club heart diamond, top right, heart club diamond, linking is heart

matrix 7 as follows,

left down, diamond heart club, bottom across heart club diamond, linking is heart,

matrix 8 as follows,

top across, diamond club heart, right down heart diamond club, linking is heart


Explained:
in total we have the following links...

3x club
2x diamond
3x heart

There needs to be one more diamond linked to complete the patterns...
Thus answer A is the only linked diamond that goes one way and then across.

Matrix A as follows,

top across, heart diamond club, center down, diamond club heart, linking is diamond

This completes the theory behind my solution.

[madhatter1979]

Regarding the initially-posted puzzle, I understand why the agreed-upon answer is correct, but I'm curious..

Spoiler:

Could D also be correct? Here's my reasoning..

The symbols in the top-left square, when the square is rotated to any degree, does not match any other pattern of symbols in any other square (there are no other squares with three hearts in a row on any side, no other square with all clubs in one "corner", and no other square with two diamonds in a middle row and one in an adjacent corner, no matter which degree they're rotated). However, if you rotate around all of the other squares, they all have at least one symbol pattern match. The same is the case with all of the possible answers except for D, and since the only other square in the puzzle that has no matches is the top-left square, I figured that the bottom-right square, rotated to any degree, should not contain a symbol pattern that matches any other square, either.

[Adam Di Giorgio]

Why complicate this? It's simple. There are three where three diamonds do not touch, three where three spades do not touch, but there are only two where three hearts do not touch. This is the problem with an IQ test, the trickery. Lots of bullshit to wade through.

[Tirian]

I hate these things, because you can find a pattern that generates any one of the possible answers, so which one is "right" is entirely arbitrary. I wouldn't mind so much if they didn't actually put them on IQ tests and such. Or at least gave you some quantitative metric by which to judge possible solutions.

The reason that SPM is an authentic measure of intelligence is that people ignore that these matrices are presented in a metacognitive context. I am confident that if we all saw the three questions before this one in the test that we would have no significant disagreement over which was the correct answer and why, because the intelligence Raven was measuring was whether you can spot a logical pattern in a simple design and apply that same pattern to a complex design. Just seeing the hardest problem in a gantlet without knowing the theme that was being developed is just putting us all in the position of guessing the biases of whoever developed the question, which is not an authentic measure of intelligence.

[Igidich]

Very good point, Tirian.

I've been studying Wittgenstein on rule-following. He concludes that there must be a sort of non-rational agreement in practices shared by humans for the notion of following a rule to make sense at all. His Philosophical Investigations is smattered with imagined scenarios called 'language games', in which such practices can be seen in simple isolation, and then built up into more complex games; all to illustrate how our agreements in practices ground our use of language. What you say reminded me of that. The 'correct answers', and the rules we are supposed to discover, are largely arbitrary when divorced from the groundwork of building up the practice of looking for rules in certain ways.

I'm sorry, maybe little of that made sense. But I'm working on my dissertation right now, and it's good to get these things off my chest, even if I am currently procrastinating.

[Stolten]

Hello

Another little psychedelic puzzle that I just created. You would probably need to see the image full size to distinguish the circles. Happy puzzling!!!

Best regards, Stolten

[laputanmachine]

Hi, i don't know if someone still checks this page, but i've got a correct answer to the first matrix, it's quite simple, though i was thinking about it for a few days:
It's really "A" because all other answers have patterns that do not appear in the "main" group, so the "difficulty" of this task really downgrades a lot, and it's up to perception. There is no obvious logic or interaction between initial elements, except repetitive patterns (sometimes unique) for each symbol, so the whole thing is like "oh, put away "cow", "tree" and "bird", because the correct answer is "Megatron" and it fits right between "Human Annihilation", "War" and "Transformers stuff" " I mean, there is diagonal row of hearts, but no diagonal row of clubs, or horizontal rows, or patterns that clubs have in "C" (because only diamonds have such pattern in initial group).

It could be really great if the arrengment of elements of the matrix was somehow isomorphic to the patterns inside these elements, and the "correct" element would itself create macro-pattern with the elements that have identical/similar patterns. I hope you know what i mean.

Sorry, if i've just repeated someone's previous answer
And sorry for bad writing/grammar/punctuation/etc. - i'm not very good at english.

[slashdottir]

I really got stuck on this puzzle when I saw it. It activated the T-Rex part of my mind that is determined to make sense of patterns.

However I think the published answer is wrong ...

Spoiler:

... and B is the correct answer.

If you look at it, the rows and columns alternate from one to the next across the whole puzzle. Then suddenly, at the last part, the symbols are rotated: clubs become hearts, hearts become diamonds and diamonds become clubs. And the only solution given that still follows the pattern is B.

Here's an illustration: http://flic.kr/p/drWFW3

[ValHarris]

Here's my answer - comments please?

Spoiler:

Label the boxes:
A B C
D E F
H I J (J is the blank)

Number the maticies:
1 2 3
4 5 6
7 8 9

Here is the pattern
The diagonal from A1 to A9 1-2-2; E1 to E9 1-2-2; that leaves answers A-B-C as possibilities

The other pattern: 4 corners in the big-box matrix - G7 and C3 are hearts; since A1 is a club, the only possible answer that fits the 4-corners pattern is answer "A" with a club in the bottom lower right corner.

[Sarek]

I think The top-left box in the question should be replaced with answer H before trying to figure this out - that's how it is in my book. But my book is first edition. I also think it makes it much easier to figure out because everyone is going to look at the first box.. and it's probably an error anyhow.
I'm not exactly wanting to 'spoil' it before anyone tries to figure it out with the first box replaced with H, but I'm going to post my solution - with spoiler tags.

Spoiler:

Heart's diamond, diamond clubs, clubbed hearts. The column increments/rotates up by 1 when passing and looping over the matrix's edge.